Central Information Systems (CIS) face a structural tension between the correctness of batch processing and the immediacy of stream processing. This paper presents a complete reference architecture for a live hybrid batch-stream system that unifies both paradigms for CIS workloads. We review the evolution from Lambda to Kappa to modern lakehouse-unified architectures, evaluate state-of-the-art technologies including Apache Kafka 3.8, Flink 1.19, Spark Structured Streaming 4.0, Kafka Streams 3.8, Debezium CDC, and open table formats (Iceberg, Delta Lake, Hudi, Paimon), and formalize consistency, fault-tolerance, and performance models. We propose a live hybrid architecture centered on a Kafka-based central event log with tiered infinite retention, a single Flink codebase with dual-mode execution (live and backfill via Iceberg-as-source using Source API), and a medallion-structured lakehouse serving layer with Redis, Druid, and vector stores. Benchmark synthesis from Halputt 2025 shows Flink achieving 1.82M events/sec with 120ms p99 latency for windowed aggregation, 2.1× Spark and 3.2× Kafka Streams, while unaligned checkpoints eliminate backpressure-induced timeouts. The architecture meets sub-500ms operational SLA and 5-15 min analytical freshness, designed for GJCST and global journal submission compliance.